W3C home > Mailing lists > Public > www-ws-desc@w3.org > September 2002

Re: Rationale for Dropping the <soap:body use=...> Attribute

From: Jacek Kopecky <jacek@systinet.com>
Date: 19 Sep 2002 14:07:19 +0200
To: ryman@ca.ibm.com
Cc: WS Description WG <www-ws-desc@w3.org>
Message-Id: <1032437239.12803.69.camel@krava>

 Arthur,
 just two points:

 1. By constraining oneself to XML Schema as the abstract type system,
one constrains oneself to the tree data model inherent in XML Schema,
other data models being out of reach (describing other data models in
XML Schema is at best a kludge). For example - what if I want to
transfer some specific RDF data in a service? How do I describe the
service using only XML Schema?
 It is true that the real representation need not be XML, but this is an
orthogonal topic.

 2. WS-I doesn't seem to support SOAP Encoding in their activities, and
if I understand you correctly, they are in fact creating their own graph
encoding. It is understandable for them, but I don't think it is
possible for WSDL 1.2 not to support SOAP Encoding properly, since SOAP
Encoding is part of SOAP 1.2 - the product of a peer W3C Working Group -
and the WS-Desc WG has sent no comments against SOAP Encoding in the
Last Call phase.

                   Jacek Kopecky

                   Senior Architect, Systinet Corporation
                   http://www.systinet.com/



On Wed, 2002-09-18 at 19:35, ryman@ca.ibm.com wrote:
> 
> Jacek,
> 
> I think it's useful to seperate the discussion into two parts:
> 
> 1) abstract (binding neutral) definition of messages in WSDL
> 2) format of messages in the SOAP binding
> 
> Concerning 1) I am in favour of just using XML schema. In fact, there is
> also discussion that the <message> element be removed and that messages be
> directly defined using schema, i.e. without <part>s. Allowing different
> schema languages is a step in the opposite direction.
> 
> Concerning 2) the WS-I.org recommendation is to just use literal. Also,
> WS-I.org is working on an algorithm to encode graphs in a way that can be
> described using a literal schema. So if the concrete message format is XML,
> then I see little benefit in allowing the concrete schema to be different
> than the abstract schema. However, there are important cases where the
> concrete message format is not XML. For example, in HTTP GET the input
> parameters are url encoded. (e.g. the input gets encoded as symbol=IBM and
> not as <symbol>IBM</symbol>). Also, if the message includes binary
> resources, then we can describe them abstractly as some restriction of
> xsd:hexBinary, but the concrete message format could be a MIME type such as
> image/jpeg using SOAP with attachments.
> 
> To summarize:
> - First, we should view the message definition as abstract and use XML
> Schema as the abstract data type language. This establishes a proper
> layering in WSDL by isolating the message definition from the bindings.
> -Second, we should define the concrete message format in the binding.
> -Third, evidence from WS-I.org tells us that for the SOAP binding, we can
> live with literal only for concrete XML messages.
> -Fourth, using literal only doesn't mean that the abstract message
> definition is always concrete since there are other important non-XML
> formats such as url encoding and MIME.
> 
> Arthur Ryman
> 
> 
>                                                                                                                                                     
>                       Jacek Kopecky                                                                                                                 
>                       <jacek@systinet.c        To:       Arthur Ryman/Toronto/IBM@IBMCA                                                             
>                       om>                      cc:       WS Description WG <www-ws-desc@w3.org>                                                     
>                                                Subject:  Re: Rationale for Dropping the <soap:body use=...> Attribute                               
>                       09/18/2002 12:11                                                                                                              
>                       PM                                                                                                                            
>                                                                                                                                                     
>                                                                                                                                                     
> 
> 
> 
>  Arhur,
>  if you want an abstract schema at the wsdl:message level, that's OK
> with me and it's understandable. On the other hand, if you want to
> remove the use attribute by saying that "literal XML Schema" is the only
> possible way in SOAP, I disagree because that either results in ugly
> *and* ambiguous data structure schemata or in disallowing other data
> models altogether (with SOAP Data Model among them).
>  I think that especially because the parts of wsdl:message should be
> described abstractly, we may need different data models right here,
> otherwise we'll say that, abstractly, WSDL only describes services that
> can transfer trees with very raw untyped references.
>  So, either let's keep use="encoded" or let's allow different schema
> languages (other than XML Schema), and I prefer the latter because it
> agrees with the requirement "abstract description of wsdl:message
> parts".
>  Best regards,
> 
>                    Jacek Kopecky
> 
>                    Senior Architect, Systinet Corporation
>                    http://www.systinet.com/
> 
> 
> 
> 
Received on Thursday, 19 September 2002 08:07:21 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:21 GMT